Local DeepSeek V4 Flash Benchmark: 24 tok/s on Epyc + RTX 5090
IntravenusDeMilo · reddit · 2026-08-24
An author shared benchmarks running DeepSeek V4 Flash (UD-Q8KXL) on a self-built "low-rent" local inference machine (AMD Epyc 7663 + 256GB ECC DDR4 + RTX 5090 32GB).
- Performance: Achieved 23.8-24.6 tokens/sec on tasks with 100-128k context.
- Latency: Prompt processing (PP) ranged from 60ms to 385ms (likely due to caching).
- Comparison: Performance improved after removing DFlash and was faster than a temporary RTX 3090 setup.
The author expressed surprise that a model of this quality runs well locally without a $10k investment.
More from Embodied
- Chinese startup DexForce trains humanoid robots with virtual data to slash costs — pstAsiatech · 2026-10-15
- EmbodiedEvo tops RoboDojo at 38.95% success by letting robots inherit failed practice — thetripathi58 · 2026-10-15
- VR dev's one line stuns a Quest 3 owner who played 3 games for 12 months ignoring mixed reality — Aiden_Tech_Ai · 2026-10-15
- ROBIB robotics meetup #3 lands in Ghent on Nov 19 with academic and industry speakers — MarwaEldiwiny · 2026-10-15
- How 3D Gaussian Splatting gives robots spatial memory of what's out of view — TheTuringPost · 2026-10-15
- RTX Spark closes the gap with M5 Macs, ushering in an "age of preference" for PC buyers — BenBajarin · 2026-10-15